Open Access
Available online http://ccforum.com/content/9/6/R623
R623
Vol 9 No 6
Research
Impact of HIV/AIDS on care and outcomes of severe sepsis
Joseph M Mrus1,2,3, LeeAnn Braun4, Michael S Yi5, Walter T Linde-Zwirble6 and
Joseph A Johnston7
1Research Physician, Health Services Research and Development, Cincinnati VA Medical Center, Cincinnati, OH, USA
2Assistant Professor, Department of Internal Medicine and Institute for the Study of Health, University of Cincinnati Medical Center, Cincinnati, OH,
USA
3Manager, Clinical Development, Infectious Diseases Medicine Development Center – HIV, GlaxoSmithKline, Research Triangle Park, NC, USA
4Associate Clinical Development Consultant, Corporate Clinical Operations, Eli Lilly and Company, Indianapolis, IN, USA
5Assistant Professor, Department of Internal Medicine and Institute for the Study of Health, University of Cincinnati Medical Center, Cincinnati, OH,
USA
6Vice President, Chief Science Officer, ZD Associates, LLC, Perkasie, PA, USA
7Clinical Research Physician, US Outcomes Research, Lilly Research Laboratories, Indianapolis, IN, USA
Corresponding author: Joseph M Mrus, joseph.m.mrus@gsk.com
Received: 27 May 2005 Revisions requested: 4 Aug 2005 Revisions received: 21 Aug 2005 Accepted: 1 Sep 2005 Published: 27 Sep 2005
Critical Care 2005, 9:R623-R630 (DOI 10.1186/cc3811)
This article is online at: http://ccforum.com/content/9/6/R623
© 2005 Mrus et al.; licensee BioMed Central Ltd.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/
2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Abstract
Introduction There has been dramatic improvement in survival
for patients with HIV/AIDS; however, some studies on patients
with HIV/AIDS and serious illness have reported continued low
rates of intensive care. The purpose of this study was to examine
patterns of care and outcomes for patients with severe sepsis
and HIV/AIDS and compare them with those of patients with
severe sepsis without HIV/AIDS.
Methods We assessed data from all 1999 discharge abstracts
from all non-federal hospitals in six US states. Patient
demographic characteristics, discharge diagnoses, resource
use, and outcomes were extracted. Analyses were performed
using chi-square, Wilcoxon rank sum, or regression techniques,
as appropriate.
Results We identified 74,020 patients with severe sepsis
(7,638 (10.3%) had HIV/AIDS) using ICD-9-CM codes.
Patients with severe sepsis and HIV/AIDS had a similar mean
length of stay (16.9 days versus 17.7 days; p = 0.0669), had
lower mean hospitalization cost ($24,382 versus $30,537; p <
0.0001), were less likely to be admitted to the intensive care unit
(37% versus 56%; p < 0.0001), and had a greater mortality
(29% versus 20%; p < 0.0001) than those without HIV/AIDS.
After adjustment for cohort differences, patients with severe
sepsis and HIV/AIDS had increased likelihood of death (OR
(95% CI) = 2.41 (2.23–2.61)) and were substantially less likely
to be admitted to the intensive care unit (OR (95% CI) = 0.54
(0.51–0.59)). When compared with those with severe sepsis
and HIV/AIDS, patients with severe sepsis without HIV/AIDS
were universally more likely to be admitted to the intensive care
unit, even when they had comorbid illnesses with equal or worse
expected in-hospital mortality (e.g., metastatic cancer).
Conclusion For patients with severe sepsis, there are
differences in care and outcomes for those with HIV/AIDS.
Further research is needed to examine the delivery of care for
patients with severe sepsis and HIV/AIDS.
Introduction
With the advent of highly active antiretroviral therapy (HAART)
in the late 1990s, opportunistic infection and mortality rates for
patients with HIV/AIDS have dramatically decreased, thus
transforming HIV/AIDS from a uniformly fatal condition to a
more manageable chronic illness [1-5]. Improvement in care
and survival have also extended to HIV/AIDS patients with
severe infections and those who receive care in the intensive
care unit (ICU) [6-9]. While studies have shown dramatic
improvement in survival related to intensive care for patients
with HIV/AIDS in the HAART era, some studies in patients
with HIV/AIDS and serious illness have reported continued
low rates of intensive care [9,10].
CI = confidence interval; HAART = highly active antiretroviral therapy; ICD-9-CM = International Classification of Diseases, 9th revision, Clinical Mod-
ification; ICU = intensive care unit; LOS = length of stay; OR = odds ratio; SS = severe sepsis.
Critical Care Vol 9 No 6 Mrus et al.
R624
In 1992, the American College of Chest Physicians/Society of
Critical Care Medicine Consensus Conference arrived at the
current definition of severe sepsis (SS) as a systemic inflam-
matory syndrome in response to infection that is associated
with acute organ dysfunction [11]. Subsequent studies have
shown that SS results in substantial morbidity and mortality for
all patients, especially for patients with comorbid illnesses,
including HIV/AIDS [12-14]. However, those data pre-date
the HAART era, and there are few data directly comparing out-
comes and resource use for patients with SS and HIV/AIDS
versus patients with SS but without HIV/AIDS. Thus, the pur-
pose of this study was two-fold: to examine patterns of care
and outcomes for patients with SS and HIV/AIDS; and to
assess differences in patterns of care and outcomes for those
with SS and HIV/AIDS versus those with SS without HIV/
AIDS.
Materials and methods
Data sources
Data from discharge abstracts for calendar year 1999 from all
non-federal hospitals from six US states (Florida, Massachu-
setts, New Jersey, New York, Virginia, and Washington) were
assessed. We selected those states based on geographic
representation, data quality and availability. Data extracted
included: patient demographic characteristics; diagnoses and
procedures (principal discharge diagnosis, up to 14 second-
ary discharge diagnoses, and hospital procedures) classified
by the International Classification of Diseases, 9th revision,
Clinical Modification (ICD-9-CM) codes; resource use (hospi-
tal length of stay (LOS), ICU use, total charges); and in-hospi-
tal mortality.
Case definition
Because no ICD-9-CM code existed at the time these data
were collected that directly identified cases of severe sepsis,
we identified cases by using an algorithm described by Angus
and colleagues [13] and adapted by others [14,15] that
required ICD-9-CM codes for a bacterial or fungal infection in
addition to acute organ dysfunction. HIV/AIDS cases were
identified using ICD-9-CM codes (042, V08) as outlined in the
Centers for Disease Control and Prevention coding guidelines
[16,17]. To improve comparability between the HIV infected
and uninfected groups, we excluded patients who were
younger than 20 years, were older than 64 years, or had preg-
nancy-related hospitalizations.
Covariate definitions
We defined a case as surgical if there was an ICD-9-CM code
for an operating room procedure other than tracheostomy.
Teaching hospital status was determined from the Health Care
Financing Administration Provider Specific File [18]. Using
classifications and methodology adapted from Deyo and asso-
ciates [19], we grouped patients into one of 10 categories
according to their pattern of chronic comorbid illnesses: no
comorbidities, HIV/AIDS, diabetes, pulmonary disease, cardi-
ovascular disease (old myocardial infarction, peripheral vascu-
lar disease, or late effects of cerebrovascular disease), renal
disease, liver disease, neoplasm (malignancy or metastatic
disease), multiple comorbidities without HIV/AIDS, and HIV/
AIDS with at least one other comorbid illness. Respiratory
infections were determined by selecting ICD-9-CM codes in
the range 460–519, and opportunistic infections were deter-
mined by selecting appropriate ICD-9-CM codes as has been
done by Keyes and coworkers [20] as well as others [21].
Outcomes
Reported outcomes were ICU use (Medical ICUs, Surgical
ICUs, or Coronary Care Units), hospital length of stay, total
cost of the admission, and in-hospital mortality. We estimated
the cost for each case by multiplying total charges by the sum
of the hospital-specific Medicare capital and operating cost-
to-charge ratios [18].
Statistical analyses
The databases were constructed in Foxpro (Microsoft Corp.,
Redmond, WA, USA) and analyses were conducted using
SAS version 8.2 (SAS Institute, Cary, NC, USA). We used chi-
square or Fisher's exact test to compare categorical charac-
teristics and Student's t test to compare continuous data.
Odds ratios (ORs) were determined using simple regression.
Adjusted analyses were performed using multivariable logistic
or linear regression, as appropriate. All available covariates
were included in the multivariable models. So that the results
would be easily interpretable, interactions among variables
were not pursued. The adjusted R2 or c-statistic is presented
for each of the models. Although distributions for LOS and
cost were not normally distributed, results were qualitatively
similar whether analyses were performed with those values log
transformed or not. Thus, we chose to not transform the data
to facilitate interpretation.
To assess the robustness of our results, we performed addi-
tional stratified analyses. Specifically, additional analyses eval-
uating mortality and ICU admission were stratified by HIV/
AIDS disease severity (presence of opportunistic infection or
not) and additional outcomes comparisons were performed
specifically with metastatic cancer (as opposed to all cancer)
diagnoses. Also, because of the imbalance in characteristics
between those with and without HIV, we were concerned
about the robustness of our multivariable results. Thus, we per-
formed additional analyses in subgroups with 'matched' char-
acteristics. Specifically, we performed two additional analyses
where we limited the cohort to patients aged 41 to 60 years
without comorbidities (other than HIV for those with HIV infec-
tion) covered by Medicaid or Medicare who were admitted to
a medical service in a teaching hospital. In the first analysis, we
assessed only those admitted with respiratory infections but
without opportunistic infections and compared outcomes for
those with and without HIV infection. In the second analysis,
we limited the analysis to only those admitted with
Available online http://ccforum.com/content/9/6/R623
R625
opportunistic infections and compared outcomes for those
with and without HIV.
Results
We identified 74,020 cases of severe sepsis, 10.3% (n =
7,638) with HIV/AIDS (Table 1). Those with SS and HIV/AIDS
were significantly younger on average (41.9 years versus 49.9
years); more likely to be male (66% versus 54%); less likely to
be white (20% versus 56%); less likely to have commercial
insurance (16% versus 42%); more likely to be admitted for
medical reasons (88% versus 69%); more likely to be admit-
ted at a teaching hospital (76% versus 61%); less likely to
have comorbid illnesses (30% versus 51%); and more likely to
have respiratory (45% versus 42%) and opportunistic infec-
tions (53% versus 9%) than those without HIV/AIDS (p ≤
0.0001 for all comparisons; Table 1).
Length of stay
For patients with SS, those with HIV/AIDS had similar mean
LOS (16.9 days) compared with those without HIV/AIDS
(17.7 days; p = 0.0669; Fig. 1). There were no significant dif-
ferences between those with and without HIV/AIDS when
LOS results were stratified by mortality. However, the impact
of HIV/AIDS on LOS varied by ICU admission status. For
patients with SS not admitted to the ICU, those with HIV/AIDS
had substantially longer LOS (15.2 days) than those without
HIV/AIDS (13.1 days; p = 0.0028), and for patients with SS in
the ICU, those with HIV/AIDS had substantially shorter LOS
(20.4 days) than those without HIV/AIDS (21.9 days; p =
0.0005). After adjusting for differences in characteristics of
patients with SS with and without HIV/AIDS through regres-
sion, those with HIV/AIDS did have a shorter LOS (-0.9 days);
however, this difference was not statistically significant (p =
0.0516; Table 2).
Hospitalization cost
A significantly lower mean hospitalization cost was observed
for patients with SS and HIV/AIDS compared with those with-
out HIV/AIDS ($24,382 versus $30,537; p < 0.0001; Fig. 1).
The cost difference between patients with SS with and
without HIV/AIDS remained significant even if results were
stratified by mortality. However, the impact of HIV/AIDS on
mean hospital cost varied by ICU admission status. For
patients with SS not admitted to the ICU, those with HIV/AIDS
incurred a similar mean cost ($18,495) to those without HIV/
AIDS ($17,615; p = 0.0755); and for patients with SS in the
ICU, those with HIV/AIDS incurred a significantly lower mean
cost ($35,594) than those without HIV/AIDS ($42,111; p <
0.0001). After adjusting for cohort differences, the difference
in hospitalization cost diminished from a difference of $6,155
to $2,706; however the difference remained statistically signif-
icant (p < 0.0001; Table 2).
Table 1
Characteristics of patients with severe sepsis
CharacteristicaHIV/AIDS No HIV/AIDS
Number of admissions 7,638 66,382
Age
Mean (SD) 41.9 (8.4) 49.9 (11.3)
Sex
Female (%) 34 46
Male (%) 66 54
Race
White (%) 20 56
Black (%) 49 18
Hispanic (%) 14 7
Other or not reported (%) 17 19
Insurance
Commercial (%) 16 42
Medicaid/Medicare (%) 48 20
Other (%)b36 38
Admission type
Medical (%) 88 69
Surgical (%) 12 31
Teaching hospital 76 61
Number of comorbiditiesc
0 (%) 70 49
≥1 (%) 30 51
Organ system failures
Respiratory (%) 35 47
Cardiac (%) 16 22
Hematologic (%) 37 25
Neurologic (%) 10 7
Renal (%) 28 24
Hepatic (%) 2 2
Number of organ system failuresd
0 (%) 0 0
1 (%) 78 78
2 (%) 17 17
≥3 (%) 5 5
Infection
Respiratory (%) 45 42
Opportunistic (%) 53 9
aThe p value for difference between patients with and without HIV/
AIDS is ≤ 0.0001 for all characteristics unless otherwise noted.
b0.0001 < p ≤ 0.0500. cNumber excludes HIV/AIDS. dp > 0.0500.
Critical Care Vol 9 No 6 Mrus et al.
R626
Intensive care unit admission and mortality
In patients with SS, those with HIV/AIDS were significantly
less likely than those without HIV/AIDS to be admitted to the
ICU (37% versus 56%; p < 0.0001) despite a statistically sig-
nificant greater overall mortality (29% versus 20%; p <
0.0001; Fig. 1). In patients with SS, those with HIV/AIDS had
significantly greater risk of death compared with those without
HIV/AIDS whether or not they were admitted to the ICU (p <
0.0001). Regardless of whether patients survived, patients
with HIV/AIDS were significantly less likely to have been
admitted to the ICU than those without HIV/AIDS (p <
0.0001). In patients with SS and HIV/AIDS, presence of
opportunistic infection did not significantly affect ICU admis-
sion rates (38% without and 36% with opportunistic infection;
p = 0.0694) or survival (29% with or without opportunistic
infection). When adjusted for age, gender, other comorbidi-
ties, race, infection site, payer type, failing organ systems,
presence of opportunistic infection, hospital teaching status,
and either ICU admission (only in mortality model) or mortality
(only in ICU admission model), patients with SS and HIV/AIDS
were more likely to die (OR (95% CI) = 2.41 (2.23–2.61))
compared with those without HIV/AIDS and were also signifi-
cantly less likely to be admitted to the ICU (OR (95% CI) =
0.54 (0.51–0.59)).
We assessed adjusted mortality and ICU admission rates for
patients with SS and comorbidities other than HIV/AIDS.
When compared with patients with SS and HIV/AIDS only
(i.e., no other comorbidities other than HIV/AIDS), patients
with SS and no cormorbidities (OR (95% CI) = 0.36 (0.33–
0.39)), or only diabetes (OR (95% CI) = 0.37 (0.33–0.42)),
pulmonary disease (OR (95% CI) = 0.38 (0.33–0.43)), cardi-
ovascular disease (OR (95% CI) = 0.39 (0.33–0.47)), or renal
disease (OR (95% CI) = 0.67 (0.56–0.80)) were significantly
less likely to die (Table 3). Those with SS and only liver disease
(OR (95% CI) = 1.28 (1.14–1.44)), only neoplasm (OR (95%
CI) = 1.79 (1.61–1.98)), or HIV with other comorbid illnesses
(OR (95% CI) = 1.67 (1.47–1.90)) were more likely to die
than those with SS and HIV/AIDS only. However, patients with
SS without HIV/AIDS were universally more likely to be admit-
ted to the ICU than patients with SS with HIV/AIDS regardless
of their comorbidities (and associated mortality rate). In an
additional comparison, we compared adjusted mortality and
Figure 1
Pattern of care and outcomes for patients with severe sepsis with and without HIV/AIDSPattern of care and outcomes for patients with severe sepsis with and
without HIV/AIDS. (a) Mean length of stay, (b) mean hospitalization
cost, (c) ICU admission rates, and (d) mortality rates are shown. Overall
results, as well as results stratified by survival and intensive care unit
(ICU) admission are shown (as appropriate). Patients with HIV/AIDS
are denoted by the white bars and patients without HIV/AIDS by the
black bars.*, p ≤ 0.0001; †, 0.0001 < p ≤ 0.05.
Table 2
Impact of HIV infection on length of stay and total cost of
admission for patients with severe sepsis
Outcome Impact of
HIV/AIDS P value Adjusted
impact of
HIV/AIDS
P value
Length of stay
(days) -0.8 0.0669 -0.9a0.0516
Hospitalization
cost ($) -6,155 <0.0001 -2,706b<0.0001
aAdjusted for mortality, intensive care unit (ICU) admission, age,
gender, comorbidities, race, infection site, payer, failing organ
systems, presence of opportunistic infection, and hospital teaching
status (adjusted R2 = 0.11). bAdjusted for mortality, ICU admission,
length of stay, age, gender, comorbidities, race, infection site, payer,
failing organ systems, presence of opportunistic infection, and
hospital teaching status (adjusted R2 = 0.64).
Available online http://ccforum.com/content/9/6/R623
R627
ICU admission rates between those with SS and HIV/AIDS
and those with SS and metastatic cancer. When compared
with those with SS and HIV/AIDS only, those with SS and met-
astatic cancer only were significantly more likely to die (OR
(95% CI) = 2.29 (2.03–2.58)) and were also significantly
more likely to be admitted to the ICU (OR (95% CI) = 1.41
(1.26–1.86)).
'Matched' analyses
To assess the robustness of our findings, we performed addi-
tional analyses in a subset of 'matched' patients. When we lim-
ited the analysis to a subset of patients aged 41 to 60 years,
without comorbidities (other than HIV for those with HIV infec-
tion), covered by Medicaid or Medicare, who were admitted to
a medical service in a teaching hospital, we obtained similar
results to the results from the whole cohort whether we
assessed patients who had respiratory infections (without
opportunistic infections) or whether we looked only at those
with opportunistic infections. In the 'matched' cohort with res-
piratory infections, those with HIV had, on average, signifi-
cantly less costly hospital stays ($2,659 less, p < 0.0001);
had shorter hospital stays (1.7 days less, p < 0.0001); were
more likely to die (OR (95% CI) = 1.86 (1.35–2.56)); and
were less likely to be admitted to the ICU (OR (95% CI) =
0.47 (0.35–0.63)). When we focused only on those with
opportunistic infections, those with HIV had, on average, sig-
nificantly less costly hospital stays ($4,490 less, p < 0.0001);
had shorter LOS (1.6 days less, p < 0.0001); and were less
likely to be admitted to the ICU (OR (95% CI) = 0.38 (0.25–
0.59)) despite similar likelihood of death (OR (95% CI) = 1.31
(0.82–2.08)).
Discussion
In this HAART-era study, we found that patients with SS and
HIV/AIDS overall had less costly hospitalizations, were less
likely to be admitted to the ICU, and had a greater in-hospital
mortality than those without HIV/AIDS. HIV/AIDS patients had
similar LOS, lower hospitalization costs, and greater mortality
than those without HIV/AIDS whether they lived, died, or were
admitted to the ICU. However, for patients with SS not in the
ICU, the trends were different. Specifically, those with HIV/
AIDS had significantly longer LOS and had somewhat higher
mean hospitalization costs (and continued higher mortality
rates) than those without HIV/AIDS. We also found that when
compared with those with SS and HIV/AIDS, patients with SS
without HIV/AIDS were universally more likely to be admitted
to the ICU, even when they had comorbid illnesses with equal
or worse expected in-hospital mortality (e.g., metastatic can-
cer). Those results were robust with qualitatively similar results
in univariate, multivariable, and subgroup analyses.
Despite having higher mortality rates, patients with SS and
HIV/AIDS were significantly less likely to be admitted to the
ICU than patients with SS without HIV/AIDS. Nicolau and col-
leagues [22] studied patients with Pneumocystis carinii pneu-
monia with and without HIV/AIDS and had similar findings.
What is unclear and cannot be discerned from our data is
whether that difference in care is inappropriate because of
physician or healthcare system bias or whether the difference
is appropriate and based on differences in patient preference
(e.g., advanced directives) or clinical differences between
patients with and without HIV/AIDS. Existing evidence sug-
gests there may be clinical biases against aggressive treat-
Table 3
Likelihood of death or ICU admission by comorbidity for patients with severe sepsis
Comorbidity Adjusted odds ratio for mortalitya (95%
confidence interval) Adjusted odds ratio for ICU admissionb (95%
confidence interval)
Only HIV/AIDS Reference group Reference group
No comorbidity 0.36 (0.33–0.39) 1.85 (1.70–2.01)
Only diabetes 0.37 (0.33–0.42) 1.82 (1.66–2.00)
Only pulmonary disease 0.38 (0.33–0.43) 1.93 (1.71–2.18)
Only cardiovascular disease 0.39 (0.33–0.47) 1.93 (1.66–2.24)
Only renal disease 0.67 (0.56–0.80) 1.76 (1.52–2.04)
Only liver disease 1.28 (1.14–1.44) 2.40 (2.16–2.67)
Only neoplasm 1.79 (1.61–1.98) 1.62 (1.48–1.79)
HIV with other comorbid illness(es) 1.67 (1.47–1.90) 1.07 (0.95–1.22)
Multiple comorbid illnesses without HIV/AIDS 0.99 (0.89–1.10) 1.75 (1.59–1.93)
aAdjusted for intensive care unit (ICU) admission, age, gender, race, infection site, payer, failing organ systems, presence of opportunistic
infection, and hospital teaching status (c-statistic = 0.80). bAdjusted for mortality, age, gender, race, infection site, payer, failing organ systems,
presence of opportunistic infection, and hospital teaching status (c-statistic = 0.79).